One-Line Summary

Agentic AI moves from chat to action: a new agent workforce platform launches, big tech doubles down on encrypted assistants and security, and Alphabet backs AI-driven drug design.

Big Tech

Meta introduces encrypted Incognito Chat for AI conversations

Meta’s new Incognito Chat is a mode for Meta AI that keeps conversations out of server logs and adds end-to-end encryption, with messages disappearing after you leave the chat; Mark Zuckerberg calls it “the first major AI product where there is no log of your conversations stored on servers,” and Meta says rollouts are coming to WhatsApp and the Meta AI app “over the coming months.” ¹

Under the hood, Meta’s Private Processing system uses trusted execution environments (TEEs), attested encrypted connections, and a stateless, forward-secure design so requests can be processed by large models without exposing content to Meta or WhatsApp. The technical white paper details components such as remote attestation (RA-TLS), oblivious HTTP, and hardware-backed confidential computing. ²

Meta AI is already embedded across Facebook, Messenger, Instagram, WhatsApp, and the Meta AI app, including a faster conversational voice mode that supports 13 languages, positioning Incognito Chat as a privacy-focused extension to experiences users can reach in their everyday apps. ³

Microsoft debuts multi-model agentic security system with top benchmark score

Microsoft introduces a multi-model, agentic security system (codename MDASH) that orchestrates 100+ specialized AI agents to audit code and found 16 new vulnerabilities across the Windows networking and authentication stack, including four Critical remote code execution issues in components like the TCP/IP stack and IKEv2. ⁴

The system identified all 21 planted flaws in a private test driver with zero false positives and achieved an 88.45% score on the public CyberGym benchmark (about five points above the next entry), plus 96% recall on five years of MSRC cases in clfs.sys and 100% in tcpip.sys. ⁴

Microsoft says MDASH is already used by internal security engineering teams and is in a limited private preview for customers, emphasizing that durable advantage comes from the agentic pipeline around models rather than any single model. ⁴

Industry & Biz

Alphabet funds AI drug design and spotlights agentic interfaces

Isomorphic Labs, an Alphabet-affiliated DeepMind spinout, raises $2.1 billion led by Thrive Capital with participation from Alphabet, Abu Dhabi’s MGX, Singapore’s Temasek, and the UK’s Sovereign AI fund to advance AI-driven drug design. ⁵

Google selects 102 startups from nearly 2,000 applicants across 16 countries for its second Gemini Startup Forum, offering hands-on sessions with Google AI teams and up to $350,000 in Google Cloud credits for eligible AI startups. ⁶

DeepMind outlines an AI-enabled pointer that lets users “point and speak” to act across apps without context-switching, and says it is bringing these principles to Chrome with pointer-based Gemini help and to a forthcoming Googlebook experience via Magic Pointer. ⁷

New Tools

BasedAI launches Hirebase, an instant AI workforce for business

Hirebase is a closed beta platform from BasedAI that lets companies deploy autonomous AI agents to execute work across everyday tools like Google Docs, Notion, Slack, WhatsApp, and Telegram, aiming to turn raw model capability into managed task execution. The launch follows BasedAI’s acquisition of Warden App’s platform IP and multi-agent orchestration stack, and the company says it has initial funding including Arche Capital. ⁸

The company pitches Hirebase to founders and operators who want to scale output without growing headcount, framing the shift as moving “from a chat experience to a doing experience” as agents research, coordinate, and execute tasks inside the tools teams already use. It is currently available via closed beta access. ⁸

Underneath, BasedAI’s vertical stack combines three layers: Hirebase for productivity workflows; BasedAPIs as an enterprise-grade, OpenAI-compatible open-weight model API layer with SLAs; and the Warden App technology to support persistent, multi-agent coordination. ⁸

What This Means for You

Agentic UX is arriving in familiar interfaces: pointing at content to “ask Gemini” in Chrome and deploying agents that act inside Slack or Docs reduces prompt writing and copy-paste overhead, which can reclaim time in research, coordination, and reporting tasks. Start with narrow, high-volume workflows where context is clear and success is easy to measure. ⁷

Privacy and control are becoming table stakes for assistants that take action on your behalf. Android’s Gemini Intelligence emphasizes explicit opt-ins, guardrails, and transparent activity logs—patterns you can mirror in internal AI guidelines and user-facing consent flows. ⁶

Security teams should note that AI isn’t just a productivity tool—it’s finding real bugs. Microsoft’s MDASH results (16 new issues, four Critical) suggest AI-assisted auditing can raise coverage while keeping false positives in check, which could change how you staff code review and triage. ⁴

If you’re building an AI product, platform incentives are growing: Google’s Startup Forum offers up to $350,000 in credits plus hands-on support, which can meaningfully lower runway costs for early-stage teams focused on AI services. ⁶

Action Items

1. Request Hirebase beta access: Apply for the closed beta and map one repetitive workflow (e.g., summarizing meeting notes into Notion) for an agent pilot.
2. Try pointer-based Gemini help in Chrome: On a product page, select items and ask Gemini to compare; note where it saves clicks versus manual prompts.
3. Review Android AI privacy controls: Read Google’s security and privacy guidance for Gemini Intelligence and draft your team’s opt-in/opt-out checklist.
4. Explore Microsoft’s MDASH preview: Ask your security lead to request the private preview and scope a limited codebase for an evaluation run.
5. Check eligibility for Google Cloud credits: If you’re an AI startup on Google Cloud, assess whether the Gemini Startup Forum benefits apply to your team this quarter.